1

私は OpenCL の初心者で、次のカーネルを動作させようとしています。構造体の配列をカーネルに渡し、その値を変更しようとしています。私の構造宣言は次のとおりです。

#define LIST_SIZE 10
#pragma pack(push, 1)

typedef struct pairt {
    int a;
    int b;
} pairt;

#pragma pack(pop)

この構造体を渡すためのバッファを作成するホスト コードは次のとおりです。

pairt p[LIST_SIZE];
p_mem_obj = clCreateBuffer(context, CL_MEM_READ_WRITE, LIST_SIZE*sizeof(struct pairt), NULL, &ret);
ret = clEnqueueWriteBuffer(command_queue, p_mem_obj, CL_TRUE, 0, LIST_SIZE*sizeof(struct pairt), &p, 0, NULL, NULL);

カーネル引数を設定するための私のコードは次のとおりです。

ret = clSetKernelArg(kernel, 0, sizeof(cl_mem), (void *)&p_mem_obj);
size_t global_item_size = LIST_SIZE;
size_t local_item_size = 2;
ret = clEnqueueNDRangeKernel(command_queue, kernel, 1, NULL, &global_item_size, &local_item_size, 0, NULL, &events[0]);
ret = clWaitForEvents(1, &events[0]);
ret = clReleaseEvent(events[0]);
ret = clEnqueueReadBuffer(command_queue, p_mem_obj, CL_TRUE, 0, LIST_SIZE*sizeof(struct pairt), p, 0, NULL, &events[1]);
ret = clWaitForEvents(1, &events[1]);

私のカーネルは次のとおりです。

struct __attribute__ ((packed)) pairt {
    int a;
    int b;
};

__kernel void simple_diff( __global struct pairt* p)
{
    int i = get_global_id(0);
    __global struct pairt *tmp = &p[i];

    tmp->a = tmp->a * -1;
    tmp->b = tmp->b * -1;

}

次の値で配列を初期化しました。

1 2
3 4
5 6
7 8
9 10
11 12
13 14
15 16
17 18
19 20

ただし、カーネルによって返される値は次のとおりです。

-298660672 -32767
0 0
-4198172 0
-298660832 -32767
-4200052 0
-1 -2
-3 -4
-5 -6
-7 -8
-9 -10

なぜこれが起こっているのか分かりませんか?

4

1 に答える 1

0

さて、私は4コアCPU上のVs2010とAMD APP SDK、および32ビットと64ビットビルドの両方のデバイスを使用したAMD6950GPUで以下のコードを試しました。私のためのその仕事...

構造体をプラグマパック1からデフォルトのパッキングに変更するか、int2ベクトルを使用してみてください

/*  
Copyright (C) Tim Child 2012
All rights reserved

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
THE SOFTWARE.

*/

#include <stdio.h>
#include <stddef.h>
#include <stdint.h>
#include <memory.h>
#include <malloc.h>

#include "CL\cl.h"

#define LIST_SIZE 10
#pragma pack(push, 1)

typedef struct pairt 
{
    int a;
    int b;
} pairt;

#pragma pack(pop)

typedef struct pair32 
{
    int32_t a;
    int32_t b;
} pair32;

void  printAligment( char * msg, size_t a, size_t b)
{
    printf("%s\n", msg);
    printf( "\t%d\t%d\n", a, b );
}

/**
    OpenCLEnvironment  An OpenCl Environment structure
**/

typedef struct OpenCLEnvironment
{
    cl_platform_id *    platformId;
    cl_uint             platformCount;  
    cl_device_id *      deviceId;
    cl_uint      *      deviceCount;    
    cl_uint             deviceTotal;
    cl_context          context;
    cl_command_queue    queue;
    cl_uint             currentPlatform;
    cl_uint             currentDevice;

} OpenCLEnvironment;

typedef struct OpenCLKernel
{
        char *      name;
        cl_program  program;
        cl_kernel   kernel;
        char *      sourceCode;
        char *      compilerOptions;
        char *      compilerErrors;

}  OpenCLKernel;

/**
    NewOpenCLEnvironment  Create a new OpenCL Environment
    @param[in]  currentPlatform  Index of the platform Id to use
    @param[in]  currentDevice    Index of the Device Id to use

    @returns OpenCL Environment
**/
OpenCLEnvironment * NewOpenCLEnvironment(cl_uint     currentPlatform, cl_uint    currentDevice)
{
    OpenCLEnvironment * environment = NULL;
    cl_int      error = 0;
    cl_uint     i = 0;
    cl_uint     sum;
    cl_uint     offset = 0;
    static cl_context_properties properties[] = { CL_CONTEXT_PLATFORM, 0, 0};

    if ( ( environment = (OpenCLEnvironment *)malloc(sizeof (OpenCLEnvironment  ) ) ) != NULL)
    {
        memset( environment, 0, sizeof(OpenCLEnvironment) );

        if ( ( error = clGetPlatformIDs( 0, NULL,  &environment->platformCount) ) == CL_SUCCESS)
        {
            environment->deviceCount = (cl_uint*)malloc( sizeof( cl_uint) * environment->platformCount  ) ;
            environment->platformId =  (cl_platform_id *)malloc( sizeof( cl_platform_id) * environment->platformCount  ) ;
            if ( environment->platformId &&
                environment->platformCount > 0 &&
                ( error = clGetPlatformIDs( environment->platformCount, environment->platformId,  0) ) == CL_SUCCESS)
            {
                sum = 0;
                for( i = 0; i < environment->platformCount; i++)
                {
                     if ( ( error = clGetDeviceIDs( environment->platformId[i], CL_DEVICE_TYPE_ALL, 0, NULL, &sum ) ) == CL_SUCCESS )
                     {
                         environment->deviceCount[i] = sum;
                         environment->deviceTotal += sum;
                     }
                }

                environment->deviceId = (cl_device_id *) malloc( sizeof(cl_device_id) * environment->deviceTotal );
                offset = 0;
                for( i = 0; i < environment->platformCount  && environment->deviceId; i++)
                {
                    if (  ( error = clGetDeviceIDs( environment->platformId[i], CL_DEVICE_TYPE_ALL, environment->deviceCount[i], 
                                        &environment->deviceId[offset],  &sum ) ) == CL_SUCCESS )
                    {
                        offset += sum;
                    }
                }
            }
        }

        if ( currentPlatform <  environment->platformCount && currentDevice < environment->deviceTotal )
        {
            environment->currentPlatform = currentPlatform;
            environment->currentDevice = currentDevice;
            properties[ 1] = (cl_context_properties)environment->platformId[ environment->currentPlatform ];
            environment->context = clCreateContext( properties, 1, &environment->deviceId[ environment->currentDevice ],
                NULL, NULL, &error );

            if (error == CL_SUCCESS)
            {
                environment->queue = clCreateCommandQueue( environment->context,environment->deviceId[ environment->currentDevice ], 0, &error);
            }
        }
    }

     return environment;
}

/**
    NewOpenCLKernel  Create a new OpenCL Kernel from Source code
    @param[in] environment  OpenC;l Environment
    @param[in] name Kernel Name
    @param[in] sourceCode  Kernel source code
    @param[in] compilerOptions command line option used to compile the  kernel

    @returns  a New OpenCL Kernel structure
**/
OpenCLKernel * NewOpenCLKernel( OpenCLEnvironment *  environment, char * name, char * sourceCode, char * compilerOptions )
{
    OpenCLKernel *  kernel = NULL;
    size_t          size = 0;
    cl_int          error = CL_SUCCESS;
    cl_int          error2 = CL_SUCCESS;

    if (  ( kernel = ( OpenCLKernel * ) malloc(sizeof ( OpenCLKernel )  ) ) != NULL )
    {
        memset(kernel, 0, sizeof( OpenCLKernel ) );
        kernel->sourceCode = sourceCode;
        kernel->name = name;
        kernel->compilerOptions = compilerOptions;

        kernel->program = clCreateProgramWithSource( environment->context, 1, &kernel->sourceCode, NULL, &error );

        error = clBuildProgram( kernel->program, 1, &environment->deviceId[ environment->currentDevice], kernel->compilerOptions,
            NULL, NULL );

        if ( error == CL_BUILD_SUCCESS )
        {
             kernel->kernel = clCreateKernel( kernel->program, name, &error );
        }
        else if ( error != CL_SUCCESS)
        {
            error2 = clGetProgramBuildInfo( kernel->program, environment->deviceId[ environment->currentDevice], 
                CL_PROGRAM_BUILD_LOG, (size_t)NULL, NULL, &size );
            kernel->compilerErrors = (char*)malloc( size  +1);

            error2 = clGetProgramBuildInfo( kernel->program, environment->deviceId[ environment->currentDevice], 
                CL_PROGRAM_BUILD_LOG, size, kernel->compilerErrors , &size );

        }

    }

    return kernel;
}


void PrintPair(char * msg, pairt * p, int n)
{
    int i = 0;

    printf("%s\n", msg);
    for ( i = 0; i < n; i++)
    {
        printf ("\t%d\t%d\n", p[i].a, p[i].b );
    }
    printf( " \n" );
}

int main ( int argc, char  ** argv)
{
    cl_int      error = CL_SUCCESS;
    cl_mem      p_mem_obj;
    size_t      global_item_size = LIST_SIZE;
    size_t      local_item_size = 2;
    cl_event    events[2];

    static pairt    unaligned[] = { 
        1, 2,
        3, 4,
        5, 6,
        7, 8,
        9, 10,
        11, 12,
        13, 14,
        15, 16,
        17, 18,
        19, 20,};

    pair32  aligned[] = { 
        1, 2,
        3, 4,
        5, 6,
        7, 8,
        9, 10,
        11, 12,
        13, 14,
        15, 16,
        17, 18,
        19, 20,};

    cl_int2 vector[] = {

        1, 2,
        3, 4,
        5, 6,
        7, 8,
        9, 10,
        11, 12,
        13, 14,
        15, 16,
        17, 18,
        19, 20,};

    OpenCLEnvironment *  environment = NewOpenCLEnvironment(0, 1);



    static char * simple_diff = "struct __attribute__ ((packed)) pairt { \
    int a; \
    int b; \
}; \
 \
__kernel void simple_diff( __global struct pairt* p) \
{                                                    \
    int i = get_global_id(0); \
    __global struct pairt *tmp = &p[i]; \
    \
    tmp->a = tmp->a * -1; \
    tmp->b = tmp->b * -1; \
\
}";

    OpenCLKernel *  simpleDiff;  

    printAligment( "Pack 1 aligned", offsetof(pairt, a), offsetof(pairt, b) );
    printAligment( "Default aligned", offsetof(pair32, a), offsetof(pair32, b) );
    printAligment( "Vector aligned", offsetof(cl_int2, s[0]), offsetof(cl_int2, s[1]) );

    simpleDiff = NewOpenCLKernel(environment,  "simple_diff", simple_diff, "" );

    p_mem_obj = clCreateBuffer(environment->context, CL_MEM_READ_WRITE, LIST_SIZE*sizeof(struct pairt), NULL, &error);
    error = clEnqueueWriteBuffer(environment->queue, p_mem_obj, CL_TRUE, 0, LIST_SIZE*sizeof(struct pairt), &unaligned, 0, NULL, NULL);

    error = clSetKernelArg(simpleDiff->kernel, 0, sizeof(cl_mem), (void *)&p_mem_obj);

    error = clEnqueueNDRangeKernel( environment->queue, simpleDiff->kernel, 1, NULL, &global_item_size, &local_item_size, 0, NULL, &events[0]);
    error = clWaitForEvents(1, &events[0]);
    error = clReleaseEvent(events[0]);
    error = clEnqueueReadBuffer(environment->queue, p_mem_obj, CL_TRUE, 0, LIST_SIZE*sizeof(struct pairt), unaligned, 0, NULL, &events[1]);
    error = clWaitForEvents(1, &events[1]);
    error = clReleaseEvent(events[1]);


     PrintPair( "Pack 1", unaligned, sizeof(unaligned)/sizeof(unaligned[0]) )
}
于 2012-07-21T02:31:48.500 に答える